Abstarct: According to the World Health Organization, heart disease is the essential reason of global deaths,. In 2008, 3 to 17 million people lost their lives due to heart disease, which covers 30% of global deaths. Cardiac signals, are the most common way of detecting heart disease and also a useful diagnostic tool for separating patients from healthy people, since it is a method with following properties, simple, inexpensive and without risk. ECG arrhythmia detection is a proper and essential method in detecting heart disease and in the development of intelligent computer systems. Accuracy of detection of QRS and R-peak are very important for applications of ECG analysises. In this thesis,we focus on extracting R-peak and the selected points on the QRS wave, using two algorithms, continuous fractional Fourier transform and geometric angle between two consecutive samples. The maximum amount of the mentioned points, is obtained at any QRS. After extracting the points and reducing the samples, extracted features are classified and evaluated using classifiers, supporting vector machine and k-Nearest Neighbor. Then the performance of the algorithms have been evaluated on MIT-BIH databaxse. The results of the classifiers are compared with recent studies. Experimental results show that the classification accuracy are more than 99%. Moreover evaluation results indicate that propose algorithms out perform the most of other methods.